Michael Galperin,
1
Abraham Nitzan,
2
and Mark A. Ratner
1
1
Department of Chemistry and Nanotechnology Center, Northwestern University, Evanston, Illinois 60208, USA
2
School of Chemistry, The Sackler Faculty of Science, Tel Aviv University, Tel Aviv 69978, Israel
(Received 17 November 2005; published 25 April 2006)
Using a model comprising a twolevel bridge connecting free electron reservoirs we show that coupling
of a molecular bridge to electronhole excitations in the leads can markedly effect the sourcedrain current
through a molecular junction. In some cases, e.g., molecules that exhibit strong charge transfer transitions,
the contribution from electronhole excitations can exceed the Landauer elastic current and dominate the
observed conduction.
DOI: 10.1103/PhysRevLett.96.166803 PACS numbers: 73.23.b, 73.50.Lw, 85.65.+h, 85.80.b
Introduction. Electron transport in molecular tunnel
junctions has been the focus of intense recent research
[14]. Theoretical modeling of tunnel conduction [5,6]
starts from Hamiltonians that contain electron transfer
(tunneling) interactions between molecule and leads as
essential elements for current transport in such junctions.
At the same time, energytransfer interactionsexcitation
(deexcitation) of the molecule accompanied by electron
hole pair annihilation (creation) in the metal are known
to strongly affect the lifetime of excited molecules near
metal surfaces [7]. An essential difference between these
interactions is that electron transfer is a tunneling process
that depends exponentially on the moleculemetal dis
tance, while energy transfer is associated with dipolar
coupling that scales like the inverse cube of this distance,
and can therefore dominate at larger distances.
How will such dipolar interactions affect the conduction
properties of molecular junctions? Here we address this
question by using the nonequilibrium Green function
(NEGF) formalism to derive an expression for the conduc
tion in a junction model that contains both electron and
energytransfer interactions, then analyze several examples
with reasonable parameters. We conclude that current
caused by electronhole excitations in the leads may be
signicant, sometimes even dominant, in situations when
strong asymmetry in the moleculelead coupling, of a
particular type explained below, is present.
As an extreme example consider the case where the
highest occupied molecular orbital (HOMO) is coupled
only to one lead, while the lowest unoccupied molecular
orbital (LUMO) is coupled only to the other. Such a
junction cannot pass current (in the absence of electronic
correlations) if the dipolar interaction is absent. However,
mixing of the HOMO and LUMO orbitals by dipolar
coupling to the leads makes conduction possible.
Realistic situations will not be that extreme; still, whenever
the HOMOLUMO transition is associated with electronic
charge transfer along the molecular bridge, we expect
some degree of such asymmetry. For example, the dipole
moment of DMEANS (4Dimethylamino4nitrostilbene)
is 7 D in the ground state and 31 D in the rst excited
singlet state [8], and 40 A
(MK)r
mm
/
j
(MK)a
m
/
m

i
mm
/
(MK)
m
=2; (10a)
(MK)<
mm
/
i
mm
/ f
K
(E)
(MK)
m
; (10b)
(MK)>
mm
/
i
mm
/ j1 f
K
(E)
(MK)
m
; (10c)
(MK)
m
2
X
kK
}V
(MK)
km
}
2
(E "
k
); (10d)
f
K
(E) jexp(E
K
)=k
B
T +1
1
(10e)
(note that the matrix
(MK)
was assumed diagonal, disre
garding possible level mixing due to coupling to the leads),
where
K
, K L; R, are the chemical potentials of the left
and right leads, respectively. This results in the standard
Landauer expression for I
et
sd
:
I
et
sd
e
@
Z
+
dE
2
X
m1;2
(ML)
m
G
r
mm
(E)
(MR)
m
G
a
mm
(E)jf
L
(E) f
R
(E)
(11)
The evaluation of I
eh
sd
is more involved. The SEs asso
ciated with the energytransfer processes can be expressed
in terms of the GFs [see Eqs. (19), (22), and (36) of
Ref. [15] ], and the selfconsistent solutions of the resulting
coupled equations are found by iterations. We use the level
populations, n
m
i
R
dE
2
G
<
mm
(E) (m 1; 2), as a test for
convergence, which is declared when the population values
at subsequent iteration steps do not change within a pre
dened tolerance, chosen below as 10
6
.
The results shown in Figs. 2 and 3 were obtained from
this procedure. A simple analytical expression for I
eh
sd
can
be obtained when
m
(L)
m
+
(R)
m
< "
21
(where
K
m
(MK)
m
+
(NK)
m
with
(NK)
m
2 Im
(NK)r
mm
) [18] and in the
PRL 96, 166803 (2006)
P HYS I CAL RE VI E W L E T T E RS
week ending
28 APRIL 2006
1668032
strong bias limit. Strong bias implies, e.g., for nega
tively biased left electrode, that
L
> "
2
and
R
< "
1
so
that f
L
1 and f
R
0 in the relevant energy range.
Under these conditions the results for the Landauer current
I
et
sd
and for the electronhole excitations induced current I
eh
sd
are obtained [15] in the forms
I
et
sd
e
@
X
m1;2
(ML)
m
(MR)
m
m
sgn(
L
R
); (12)
I
eh
sd
e
@
B
(ML)
2
(MR)
1
2
(
L
R
)
(ML)
1
(MR)
2
2
(
R
L
)
; (13)
where (x) 1 for x >0 and 0 for x <0, and where B
B
(L)
+B
(R)
is the function
B
(K)
(!;
K
) 2
Z
dE
X
kk
/
K
}V
(NK)
kk
/
}
2
(E "
k
)
(E +! "
k
/ )f
K
(E)j1 f
K
(E +!)
2}V
(NK)
}
2
eh
K
(!) (14)
evaluated at ! "
2
"
1
. Here
eh
K
(!) is density of
electronhole excitations in the lead K. Equations (12)
and (13) show that the magnitude of I
et
sd
depends on the
product
(ML)
m
(MR)
m
(m 1; 2), while I
eh
sd
is determined by
the mixed products
(ML)
2
(MR)
1
or
(ML)
1
(MR)
2
. The relative
contribution of the latter current component will be sig
nicant in asymmetric coupling situations, e.g., for
L
>
R
when
(ML)
2
>
(MR)
2
and/or
(MR)
1
>
(ML)
1
. Note that,
since
(ML)
2
(MR)
1
can be different from
(ML)
1
(MR)
2
, }I
eh
sd
}
can be asymmetric to bias reversal (see also Fig. 2). Below
we compare the magnitude of the two contributions to the
current for different junction parameters.
Numerical results. In the calculations reported below
we used the following standard choice of parameters:
T 300 K, "
1
0 eV, "
2
2 eV, and
(M)
1
(M)
2
0:2 eV. Values of other parameters are indicated in the
gures. The Fermi energy is taken at the midpoint of the
HOMOLUMO gap and the electrochemical potentials in
the left and right leads are assumed to shift with the voltage
bias symmetrically relative to this point. Numerical inte
gration was done on the energy grid spanning range from
3 to 5 eV using integration step 10
3
eV.
Figure 2 depicts the currentvoltage characteristic of the
junction for the cases of symmetric and asymmetric cou
pling between the leads and the molecular LUMO. Shown
are the total current and its two components. The following
points are noteworthy: (1) In the symmetric case the cur
rent is dominated by the usual electron and/or hole trans
port through the LUMOand/or HOMO, respectively, and is
symmetric with respect to voltage reversal. (2) The asym
metric case shows a signicant contribution of the current
associated with electronhole excitations when the LUMO
is coupled more strongly to the negatively biased electrode.
Indeed, I
eh
sd
is expected to be pronounced when the LUMO
is populated and the HOMO is empty, which happens at
such bias. (3) The asymmetric LUMO coupling leads to
asymmetry in I
eh
sd
and consequently in the total current with
respect to bias reversal.
Figure 3 shows the results of a model study of the
dependence of the sourcedrain current on the molecule
lead distance R. These results are obtained using
(MK)
m
A
(MK)
m
expj
(MK)
m
R, which accounts for the tunneling
nature of electron transfer, and taking B
(K)
(K)
=R
3
,
which reects a dipolar distance dependence of the energy
transfer to the leads [7]. The parameters used are A
(ML)
1
A
(MR)
1
0:27 eV, A
(ML)
2
0:52 eV, A
(MR)
2
0:027 eV,
(MK)
m
1
A
1
, and
(K)
0:01 eVA
3
(K L; R and
m 1; 2). The choice of A
(MK)
m
corresponds to a total life
time broadening for electron transfer into the electrodes of
0.2 eV at a distance (from each electrode) of 1 A
. The
choice of
(K)
corresponds to taking B
(K)
0:01 eVat this
distance. Both choices reect the orders of lifetimes ob
2
1
0
1
( 10
5
)
I
s
d
(
A
)
6 4 2 0 2 4 6
e (eV)
1
0
1
( 10
5
)
I
s
d
(
A
)
(a)
(b)
FIG. 2 (color online). The sourcedrain current I
sd
vs applied
voltage . Shown are the total current I
sd
(solid red line) and its
components due to direct electron transfer I
et
sd
(dashed green line)
and electronhole excitations I
eh
sd
(dotted blue line), for the sym
metric
(ML)
1;2
(MR)
1;2
0:1 eV (a) and asymmetric
(ML=R)
1
0:1 eV,
(ML)
2
0:19 eV, and
(MR)
2
0:01 eV (b) cases.
PRL 96, 166803 (2006)
P HYS I CAL RE VI E W L E T T E RS
week ending
28 APRIL 2006
1668033
served for the corresponding processes for molecules ad
sorbed on metal surfaces.
The results shown in Fig. 3 demonstrate an important
aspect of the distance dependence of the two contributions
to the total current. Obviously, both I
et
sd
and I
eh
sd
involve
electron transfer between molecule and leads, therefore
both drop exponentially with increasing tunneling length.
Their relative importance, however, depends on the details
of the moleculeleads couplings. It is easily seen from
Eqs. (12) and (13) that I
eh
sd
=I
et
sd
B=. For small and
intermediate distances, where is dominated by the elec
tron tunneling process, this ratio increases with increasing
distances. (Note that this limiting behavior is obtained only
when both left and right moleculemetal couplings de
crease together. Experimentally, one of these distances
can be controlled by moving a tip, while the other can be
changed by adding insulating layers between molecule and
substrate [19,20].) Detailed calculation shows that this
ratio can in fact become substantially larger than 1 at large
moleculelead distances, as seen in Fig. 3. In this case the
current is dominated by the contribution associated with
the electronhole excitation process.
Conclusion. We have studied, within a simple model,
the effect of dipolar energytransfer interaction between
molecule and leads on molecular conduction. We found
that such interaction, which leads to electronhole excita
tions in the contacts, can affect the currentvoltage char
acteristic of the junction in a substantial way and cannot in
general be disregarded. The contribution of this interaction
can dominate the overall conduction for particular asym
metric coupling where the molecular LUMO and/or
HOMO are coupled differently to different leads. In addi
tion, because of the different dependence of electron and
energy transfer on the moleculeleads distance, the relative
importance of I
et
sd
and I
eh
sd
depends on this distance, and can,
in some cases, result in strong dominance of I
eh
sd
at large
moleculelead separations.
We thank the NSFNNI program, the DARPA MolApps
initiative, and the Durint/MURI program of the DOD for
support. A. N. thanks the Israel Science Foundation and the
U.S.Israel Binational Science Foundation for support.
[1] Molecular Electronics II, edited by A. Aviram, M. Ratner,
and V. Mujica (The New York Academy of Sciences,
New York, NY, 2002), Vol. 960; Molecular Electronics
III, edited by J. R. Reimers, C. A. Picconatto, J. C.
Ellenbogen, and R. Shashidhar (The New York
Academy of Sciences, New York, NY, 2003), Vol. 1006.
[2] A. Nitzan, Annu. Rev. Phys. Chem. 52, 681 (2001).
[3] J. R. Heath and M. A. Ratner, Phys. Today 56, No. 3, 43
(2003).
[4] Molecular Nanoelectronics, edited by M. A. Reed and
T. Lee (American Scientic Publishers, Stevenson
Ranch, CA, 2003).
[5] A. Nitzan and M. A. Ratner, Science 300, 1384 (2003).
[6] S. Datta, Quantum Transport: Atom to Transistor
(Cambridge University Press, Cambridge, England, 2005).
[7] R. Chance, A. Prock, and R. Silbey, Adv. Chem. Phys. 31,
1 (1978).
[8] S. N. Smirnov and C. L. Braun, Rev. Sci. Instrum. 69, 2875
(1998).
[9] V. L. Colvin and A. P. Alivisatos, J. Chem. Phys. 97, 730
(1992).
[10] M. Galperin and A. Nitzan, Phys. Rev. Lett. 95, 206802
(2005).
[11] Y. M. Niquet et al., Phys. Rev. B 65, 165334 (2002).
[12] L. V. Keldysh, Sov. Phys. JETP 20, 1018 (1965).
[13] Y. Meir and N. S. Wingreen, Phys. Rev. Lett. 68, 2512
(1992).
[14] N. E. Bickers, Rev. Mod. Phys. 59, 845 (1987).
[15] M. Galperin and A. Nitzan, condmat/0602217.
[16] G. D. Mahan, ManyParticle Physics (Kluwer Academic
and Plenum, New York, 2000), 3rd ed.
[17] H. Haug and A.P. Jauho, Quantum Kinetics in Transport
and Optics of Semiconductors (Springer, Berlin, 1996).
[18] Explicit expressions for
(N)
m
are obtained in this limit in
terms of the level populations n
m
(m 1; 2) [see Eq. (37)
of Ref. [15] ], which in turn are obtained from the self
consistent calculations.
[19] N. A. Pradhan, N. Liu, and W. Ho, J. Phys. Chem. B 109,
8513 (2005).
[20] J. Repp, G. Meyer, S. M. Stojkovic, A. Gourdon, and
C. Joachim, Phys. Rev. Lett. 94, 026803 (2005).
2 4 6 8 10 12 14
R (Angstrom)
10
11
10
10
10
9
10
8
10
7
10
6
10
5
I
s
d
(
A
)
FIG. 3 (color online). The sourcedrain current I
sd
at voltage
3 V vs moleculeleads distance R. See text for the choice of
coupling parameters. Shown are the total current I
sd
(solid red
line) as well as its I
et
sd
(dashed green line) and I
eh
sd
(dotted blue
line) components.
PRL 96, 166803 (2006)
P HYS I CAL RE VI E W L E T T E RS
week ending
28 APRIL 2006
1668034